search

GPflow / GPflowOpt

Bayesian Optimization using GPflow
Apache License 2.0
270 stars 61 forks source link

Discrete variables optimization #122

Open HolmKiilerich opened 3 years ago

HolmKiilerich commented 3 years ago

Hey there, I found your project which seems promising for a problem, I am currently working on. However, I as far as I can see, the option to include discrete variables in the optimization is not yet implemented? Is this correct? and if so, are there any developement in this direction currently going on?

sedenhazal commented 3 years ago

Hi,

I have the same issue. Instead of using gpflowopt.domain.ContinuousParameter, I thought we can use gpflowopt.domain.Parameter for defining discrete variables. However, I could not understand how to use it.

icouckuy commented 3 years ago

If the discrete variables are ordinal then you can treat them as continuous parameter for the modeling. When optimizing the acquisition function you can just evaluate the discrete parameter on a grid (if there are not that many values) or just round the optimum found by the scipy optimizer.